During a bus-commute I happened to consider sentences where case assignment conflicts may appear. Take, for instance:
Did you know who appeared then?
Who can be seen as two things here - the subject of
appear or the object of
know (in English, I am pretty sure it is the subject of appear, and not the object of 'know'; consider what happens to the meaning of the entire sentence if you substitute 'who', with, e.g. 'John' - the element being questioned (and therefore the underlying structure) changes entirely. This does not guarantee that other languages do not have it occupy slots that are in a more clear relation to one of the verbs or the other - there probably even are languages where which way to go is free. To be clear, my contention here is that in English, 'who appeared then' is a constituent under 'know', where 'who' also kind of works as a CP for all of 'who appeared then', telling us the kind of role that the subclause has. In this kind of situation, it'd be interesting to consider how different case-attributing strategies can come in conflict, and how that conflict is resolved.
First, I will look into the ergative situation. One simple approach would be to favor intransitives and antipassives in the nested clause, thus making the object agree in case with the expected role it has in the nested clause. However, that is somewhat boring - we could also simply force it to be absolutive without restricting the verb in any way, and present a situation where absolutives explicitly can be subjects of nested transitive verbs. We could also go for a markedness-approach, and permit the noun to be ergative, thus enabling a kind of ergative-marked object.
Another possibly interesting approach would be to have a language that combines nom-acc and erg-abs approaches, and to some extent resolves conflicts by defaulting to abs in the previously described way, overruling any other case assignments that are called for in either of the clauses.
If we skip the ergative approach (ergativity is so previous decade anyway), there's also the possibility that accusative would beat nominative, because it's more
marked. (In the sense that it is slightly more remarkable for a noun in general to be in the accusative than in the nominative, for a somewhat specific sense of
remarkable.)
Finally, it seems it is not unusual for languages with case markings to have a 'wastebasket case' for whenever the speakers are uncertain which case to use. In English, this apparently is fairly certainly the oblique case (me, you, him, her, ...) although some are artificially overdoing the nominative as the wastebasket (thus producing stuff like 'they never tell you and I of it'). This seems to be the kind of situation where wastebasket use could be called for. The wastebasket need not be nominative or accusative though - it could conceivably be somewhat less of a core case - I'd argue genitives, datives, partitives and instrumentals, depending on their use in the language, may all make sense for wastebasket case - provided their use is varied enough in the language. The sentence could end up something like 'do you know by who arrived' or 'do you know of who arrived'?
Ultimately though, for a wastebasket case to enter into that kind of structure, this particular construction has to be fairly recent, and some other way of expressing the same notion must be falling out of use - possibly something with noun-verbs or participles or such, i.e.
do you know whose arrival then?
